B21B-01 INVITED
Modeling Carbon Cycles and Trace Gas Budgets of Savannas Worldwide using NASA Satellite Products: Focus on Climate and Land Use Change Impacts
We are evaluating results from the ecosystem model called NASA-CASA (Carnegie-Ames Stanford Approach) for characterizing carbon cycles and trace gas budgets for savannas on a global scale. The NASA-CASA model has been formulated to run at monthly time intervals for the years 2000 to the present using NASA satellite data inputs from the MODIS and CERES sensors. Maps of net primary production (NPP) and net ecosystem exchange of carbon from the NASA-CASA model show the savanna regions of the globe where climate and land cover conversions have had the greatest impacts over the past 5 to 20 years. Special emphasis will be focused on the Brazilian Cerrado region as savanna ecosystems undergoing rapid transitions to intensive agricultural land uses. http://geo.arc.nasa.gov/sge/casa/
B21B-02 INVITED
What drives spatio-temporal variability of Africa's carbon fluxes and their connection to climate oscillations?
Terrestrial process models and atmospheric inversions attribute a large fraction of global interannual variability in net ecosystem carbon dioxide exchange to tropical lands. Africa, one of the least well-understood components of the global picture, reportedly contributes as much as one third of the year-to-year variation in global net carbon dioxide source/sink dynamics. Surprisingly small interannual variability of African fire emissions means photosynthesis and respiration can be expected to play a particularly important role in governing the continent's interannual variability of net ecosystem exchange. In this presentation, we report results from a biophysical model used to study what drives spatio-temporal patterns of photosynthesis and respiration in the various eco- climatic settings across the African continent, providing in-depth analysis of continental-scale patterns and focusing more specifically on what causes their year to year variation. Hydrologic fluctuations are the most important drive of interannual variability in photosynthesis since water deficit is prevalent and interannual variability of rainfall is high. While hydrologic fluctuations are often erratic, rainfall across much of Africa resonates with the El Nino Southern Oscillation and the Indian Ocean Dipole and this drives sizeable anomalies of the gross fluxes with substantial temporal correlation. In addition, we use flux tower observations to study where models are likely to suffer deficiencies associated with complicated biologically-regulated lags triggered by discrete wetting events.
B21B-03
Influence of Vegetation Cover on Rain Pulse Responses in Semi-Arid Savannas in Central Texas
Savannas in central Texas are dominated by live oak (Quercus virginiana) and Ashe juniper (Juniperus asheii) underlain by perennial, C3/C4 grasslands, and are increasingly becoming juniper and mesquite dominated due to overgrazing and suppression of wildfires. Since 2004, we have been investigating how carbon, water and energy exchange in these rain-limited savannas respond to rainfall variability and this observed vegetation change. In semi-arid regions, rainfall pulses provide inputs of soil moisture and trigger biotic activity in the form of plant gas exchange and microbial metabolism as well as water dependent physical processes in the soil. Each of these components has a different characteristic response curve to soil moisture and integrates soil water content over a different range of depths. Here we focus on examining how the observed increase of woody species in central Texas savannas alters the response of net ecosystem exchange and its components, ecosystem respiration and gross ecosystem exchange, to rain pulses. Using data we have collected over the last three years from three Ameriflux tower sites at Freeman Ranch near San Marcos, TX (C3/C4 grassland, juniper/mesquite savanna with 50 percent woody cover, and oak/juniper woodland), we quantify the responses of both ecosystem respiration and daily carbon uptake to rainfall pulses throughout the year. Specifically, we look at the enhancement and persistence of ecosystem respiration and carbon uptake responses following a pulse, and isolate the main controlling factors on the observed response: seasonality, antecedent soil moisture and temperature, or previous pulses. In all three land covers, the general response to precipitation pulses is a respiration pulse followed by an increase in total carbon uptake. Differences in pulse responses observed at the savanna site compared to the grassland and woodland sites can be explained, in part, by the observed differences in rooting structure and photosynthetic capacity due to differences in plant functional groups and leaf area index. The woodland site is most sensitive to winter pulses in terms of enhanced sink strength following pulses and is dependent on both temperature and pulse size. Both the grassland and shrubland sites show greater sink strength following summer pulses, rather than winter pulses. Both the ecosystem respiration and net uptake responses in all three sites are dependent upon whether there was shallow versus deep soil moisture recharge following pulses. Implications for the influence of future climate change on carbon dynamics in these savanna ecosystems will be discussed.
B21B-04
Changes in Carbon Emissions in Colombian Savannas Derived From Recent Land use and Land Cover Change
The global contribution of carbon emissions from land use dynamics and change to the global carbon (C) cycle is still uncertain, a major concern in global change modeling. Carbon emission from fires in the tropics is significant and represents 9% of the net primary production, and 50% of worldwide C emissions from fires are attributable to savanna fires. Such emissions may vary significantly due to differences in ecosystem types. Most savanna areas are devoted to grazing land uses making methane emissions also important in savanna ecosystems. Land use change driven by intensification of grazing and cropping has become a major factor affecting C emission dynamics from savanna regions. Colombia has some 17 MHa of mesic savannas which have been historically burned. Due to changes in market demands and improved accessibility during the last 20 years, important areas of savannas changed land use from predominantly extensive grazing to crops and intensive grazing systems. This research models and evaluates the impacts of such land use changes on the spatial and temporal burning patterns and C emissions in the Orinoco savannas of Colombia. We address the effects of land use change patterns using remote sensing data from MODIS and Landsat, ecosystem mapping products, and spatial GIS analysis. First we map the expansion of the agricultural frontier from the 1980s-2000s. We then model the changes in land use from the 1980s using a statistical modeling approach to analyze and quantify the impact of accessibility, ecosystem type and land tenure. We calculate the effects on C emissions from fire regimes and other sources of C based on patterns and extent of burned areas in the 2000s for different savanna ecosystem types and land uses. In the Llanos the fire regime exhibits a marked seasonal variability with most fire events occurring during the dry season between December-March. Our analysis shows that fire frequencies vary consistently between 0.6 and 2.8 fires.yr-1 per 2,500 Ha among the different savanna ecosystem types. Highest frequencies and largest burned areas occur in the less accessible well-drained savannas of the southern part of the region. The analysis also reveals a close relationship between land tenure and fire regimes, with highest frequencies in Indigenous Reserves, followed by private land ranches and National Parks, indicating that most fires are human induced. By 2000 more than 500k hectares of natural savannas were transformed to sown pastures (Brachiaria spp.), and some 100k hectares were planted with oil palm and irrigated rice. Such changes have taken place in more accessible areas and slightly better soils. In areas subject to land use change and intensification a significant reduction in fire frequency can be observed. Because such land use changes have been occurring in savanna types with better soils and higher aerial biomass values, the average effect on reduction of C-emissions is some 30 to 50% larger than the effect on fire area reduction. Our results indicate a reduction of fire frequencies greater than 80% in areas where savannas were replaced by introduced Brachiaria pastures. However the reduction in C emissions from fire reduction in these pastures is exceeded by the parallel emissions from the increase in the cattle stocking rates with a net effect of an additional emission of 0.5 Gt.CO2 equivalents. We make preliminary projections of future emission trends based on the land use change model, and we discuss the likely effects of future sources and sinks of C expected from the increase of irrigated rice crops and from projected oil palm and timber plantations.
B21B-05
Land Use and Changes in Carbon Budget in the Brazilian Cerrado
Tropical savannas cover 22.5 x 106 km2, an area nearly 30% larger than the area of tropical forests. Although the average carbon \(C\) content of savanna vegetation is only about 25% as great as tropical forest vegetation (29 vs. 120 Mg C ha-1), land use changes in tropical savannas are even more rapid than changes in tropical forests. The Brazilian savanna, locally known as Cerrado, covers about 2 x 106 km2 and is the largest savanna formation in South America. Its area is comparable to the Miombo savanna of Southern Africa. Biomass in the Cerrado varies from 1.9 Mg C ha-1 \(grassy campo limpo\) to 30.5 Mg C ha-1 \(woody cerradão\). Cerrado vegetation can be highly productive; annual net ecosystem exchange fluxes as high as 2.5 Mg C ha-1 yr-1 have been measured although lower values are more common. Assuming approximately 40 years of land use conversion and an average net biomass change \(29 Mg C ha-1\), this would lead to an average loss of C from the Cerrado of nearly 0.1 Pg C y-1. These values can be higher if belowground biomass is included as in Cerrado the ratio of belowground to aboveground biomass reaches values as high as 7.7. Fire is a principle factor controlling vegetation dynamics in the Cerrado \(especially the ratio of grass to woody biomass\). Frequent fires kill trees and shrubs favoring grasses favoring more open vegetation types. Inversely fire suppression favors woody growth. Advances in agricultural productivity have made the Cerrado the leading region of Brazil for beef cattle production and soybean production. It is estimated that between 40% and 55% of the region has been converted to pasture and other agricultural uses with peak rates in the early 1970's. Increasing international demand of biofuels represents a new aspect of land use in the region. Soil organic matter stocks exceed biomass stocks and data on soil C storage with conversion of native savanna into pasture indicated that well-managed, cultivated pastures may provide enough C input to maintain or even slightly increase soil C contents under native vegetation. However, C input from degraded pastures may be too low to sustain the high soil C storage under native Cerrado. Soil C accumulation under pastures over the previous native stocks only occurs with nutrient inputs through fertilization and legumes. The magnitude of the accumulation is low compared to the initial stocks under native Cerrado \(100 Mg C ha-1 for 100 cm soil depth\). Model simulation of C and N stocks under soybean with millet as cover crop or maize as a second crop were comparable to those under native vegetation. In contrast, soil C and N contents under soybean monoculture with bare fallow were simulated to decline with approximately 30% after 30 years. Considering the conversion of native Cerrado to agricultural land a yearly C input of about 8.5 Mg C ha-1 yr-1 was necessary to maintain the initial soil C levels under the native savanna. These C inputs include both crop residues and roots from soybean and millet or maize. The C input under the soybean-fallow system \(~ 4.2 Mg C ha-1 yr-1\) was insufficient to sustain these C levels. As indicated for pastures, long-term accumulation of soil C can only be expected when the net N balance of the cropping systems is positive.
B21B-06 INVITED
Quantifying Carbon in Savannas: The Role Of Active Sensors
Savannas worldwide are experiencing rapid changes as a consequence of anthropogenic activity, climatic alteration and natural events. Such changes are impacting upon the carbon cycle, with both gains (e.g., through regrowth and woody thickening) and losses (e.g., through deforestation, dieback or burning) occurring. At local to regional scales, optical data from airborne and/or spaceborne sensors (e.g., aerial cameras, Landsat, MODIS) have remained a primary source for characterising savannas and quantifying changes in extent, condition and productivity. However, such data have been limited in quantifying the woody components and changes in these as a function of tree growth or mortality. For these reasons, regional estimates of standing carbon stocks and changes in these over time have also proved difficult to generate. In recent years, technological advances in Synthetic Aperture Radar (SAR) and Light Detection and Ranging (LiDAR) have provided increased opportunities for quantifying the biomass and structure of woody vegetation at local to regional scales. For retreiving biomass, empirical relationships with SAR backscatter have been limited because of saturation of the signal. However, many wooded savannas support a relatively low biomass and their comparative openness leads to a greater diversity of scattering processes between the vegetation components (e.g., branches, trunks) and the ground surface. More information on the forest volume can therefore be extracted which, in turn, has led to the development of new algorithms (e.g., parameter estimation, inversion) for retrieving the biomass and also structure (e.g., stem density) of savannas from SAR, including those that overcome saturation. Structural measures relating to biomass (e.g., vegetation height) have further been retreived using interferometric SAR (InSAR) and polarimetric-interferometry (PolInSAR). In many cases, algorithm development has been advanced using simulation models. LiDAR has provided complementary information on the vertical and horizontal distribution of plant elements within the forest volume and spaceborne data from the ICESAT Geoscience Laser Altimeter System (GLAS) in particular have shown promise for regionally estimating vegetation height in savannas. The use of airborne LiDAR has largely been restricted to local areas because of cost and platform availability. Nevertheless, these and other airborne (e.g., hyperspectral) datasets have played a pivotal role in providing tree to stand level information that support the interpretation of spaceborne LiDAR and SAR data and development of algorithms for biophysical parameter retrieval. Whilst SAR, LiDAR and optical sensors can separately provide unique information on savannas, the integration of data and products from each has provided the greatest opportunities for quantifying the distribution and dynamics of carbon within savannas. Examples include mapping of the growth stage of regenerating vegetation, discriminating forest types, detecting dead standing timber, and quantifying changes in land cover and forest condition. The resulting datasets have provided additional input to carbon models in savanna regions.
B21B-07
Modeling Radiation and Photosynthesis of a Heterogeneous Savanna Woodland Landscape with a Hierarchy of Model Complexity
Simple but realistic modeling of radiation transfer within heterogeneous canopy has been a challenging research question for decades and is critical for predicting ecological processes such as photosynthesis. The Markov model proposed by Nilson (1971) is theoretically sound to meet this challenge. However, it has not been widely used because of the difficulty of determining the clumping factor. We propose an analytical approach to calculate clumping factors based on the average characteristics of vegetation distributed across a landscape. In a savanna woodland in California, we simulate the photosynthesis of the landscape in three different ways: 1) the crown envelope and location of each tree is spatially-explicitly specified, 2) the canopy is assumed to be horizontally homogeneous within which leaves are randomly dispersed as a Poisson process, and 3) the canopy is horizontally homogeneous but leaves are clumped and distributed with a Markov process. We find that the Markov model can achieve much better performance than the Poisson model. Incorporating the clumping factor can reduce the percent error of CO2 assimilation estimation significantly (e.g., from around 50% to 10% when local leaf area index is 4.5 m2/m2). The results indicate that our approach of calculating clumping factors has applications in terrestrial ecosystem modeling, particularly where accurate representation of "system heterogeneity" (e.g., savannas and woodlands) is required.
B21B-08 INVITED
Multi-Sensor Model-Data Assimilation for Improved Modeling of Savanna Carbon and Water Budgets
Model-data assimilation methods are increasingly being used to improve model predictions of carbon pools and fluxes, soil profile moisture contents, and evapotranspiration at catchment to regional scales. In this talk, I will discuss the development of model-data assimilation methods for application to parameter and state estimation problems in the context of savanna carbon and water cycles. A particular focus of this talk will be on the integration of in situ datasets and multiple types of satellite observations with radiative transfer, surface energy balance, and carbon budget models. An example will be drawn from existing work demonstrating regional estimation of soil profile moisture content based on multiple satellite sensors. The data assimilation scheme comprised a forward model, observation operators, multiple observation datasets and an optimization scheme. The forward model propagates model state variables in time based on climate forcing, initial conditions and model parameters and includes processes governing evapotranspiration, water budget and carbon cycle processes. The observation operators calculate modeled land surface temperature and microwave brightness temperatures based on the state variables of profile soil moisture and soil surface layer soil moisture at less than 2.5 cm depth. Satellite observations used in the assimilation scheme are surface brightness temperatures from AMSR-E (passive microwave at 6.9GHz at horizontal polarization) and from AVHRR (thermal channels 4 & 5 from NOAA-18), and land surface reflectances from MODIS Terra (channels 1 and 2 at 250m resolution). These three satellite sensors overpass at approximately the same time of day and provide independent observations of the land surface at different wavelengths. The observed brightness temperatures are used as constraints on the coupled energy balance/microwave radiative transfer model, and a canopy optical model was inverted to retrieve leaf area indices from observed reflectances in optical wavebands. Results show that the multiple constraints approach is effective in identifying and reducing the influence of bias on the resultant analysis that occurs when only single observation data sets are used. Reductions in error and bias lead to improved prognoses of soil profile water store and forecasts of rainfall runoff. The development and routine application of model-data assimilation methods in savanna biophysical modeling will improve performance of ecosystem biophysical models, assist with the design of filed campaigns to maximize uncertainty reduction, fill gaps in knowledge of the carbon and water dynamics of savannas and provide better information on which to base decision making to solve natural resource management problems in this biome.